[Piglit] [PATCHv2 0/5] arb-enhanced-layouts: explicit-offset piglits

[Timothy Arceri]

On Mon, 2015-11-02 at 16:40 +0000, Emil Velikov wrote:
> On 2 November 2015 at 16:20, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> > On Mon, Nov 2, 2015 at 11:14 AM, Emil Velikov <emil.l.velikov at gmail.com>
> > wrote:
> > > On 1 November 2015 at 22:59, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> > > > On Sun, Nov 1, 2015 at 5:53 PM, Timothy Arceri
> > > > <timothy.arceri at collabora.com> wrote:
> > > > > On Tue, 2015-10-27 at 15:34 +0000, Emil Velikov wrote:
> > > > > - Your a bunch or your tests require the extension and also the glsl
> > > > > verion to
> > > > > be 4.30 where we know the extension will exist. You should instead
> > > > > change the
> > > > > glsl version to the lowest version that the extension requires
> > > > > otherwise these
> > > > > test will not be run on Mesa until a driver enables 4.30.
> > > > 
> > > > Actually the functionality isn't available until GLSL 4.40. However
> > > > the ext spec says:
> > > > 
> > > >     OpenGL 3.1 and GLSL 1.40 or later are required.
> > > > 
> > > > So that should be the requirement of the tests as well, unless they're
> > > > also testing things that only came about in later GL versions (e.g.
> > > > GS).
> > > Actually I've added the 4.30 requirement is due to the std430 layout
> > > qualifier. Having a closer look I cannot quite find the specific
> > > extension that introduces it. Does anyone have any tips so we can add
> > > it to the list, and drop the version as suggested (to 1.40).
> > 
> > SSBO introduces layout(std430) afaik. According to the spec, it
> > requires GL 4.0, but... meh.
> > 
> I'm fine either way bth. As long we have some consensus -> 1.40 +
> ssbo, 4.00 + ssbo or other.

I would be happy with 1.40 + ssbo the SSBO piglit tests use std430 with the
following combo so 1.40 + ssbo should be fine:

	#version 130
	#extension GL_ARB_shader_storage_buffer_object : require
        #extension GL_ARB_uniform_buffer_object : require

> 
> > > 
> > > It just hit me that despite the "required" extensions at the top, I'm
> > > only using 'enable' in the actual tests. Upon closer look some
> > > piglits, do the same while others, us required in both places. Any
> > > suggestions which one we should opt for ?
> > 
> > glslparsertest/shader_runner will skip if those exts aren't available.
> > Given that they are, require/enable don't make much of a difference.
> > 
> Having another look at glslparsertest/shader_runner neither of which
> parses the config section. Is that done by the python framework ?
> 
> Thanks
> Emil